Can artificial intelligence infringe copyright? Some reflections

Creative machines consume. They often devour huge amounts of data as part of learning processes including books, photographs, images, articles, social media feeds, videos and other kinds of content - data are the building blocks of algorithmic creativity. Programs that generate music, for example, are fed with huge quantities of source material, from hits at the top of the charts to lesser-known pieces, which they assess with a view to finding patterns. This inevitably raises risks of copyright infringement - both in relation to the inputs and outputs of the AI system - as a substantial amount of the data used may be subject to protection. Who should be liable for infringement? Are any exceptions to infringement available in these circumstances? Should we avoid a binary regime whereby algorithmic creativity is given a more generous fair use or fair dealing treatment? In this contribution, we explore the above questions with a focus on key jurisdictions including the UK, EU and the United States.